Research map / Detection and segmentation / Open-set and few-shot detection
Deep Contextual Attention for Human-Object Interaction Detection
Tiancai Wang, Rao Muhammad Anwer, Muhammad Haris Khan, Fahad Shahbaz Khan, Yanwei Pang, Ling Shao, Jorma Laaksonen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add context-aware appearance features (ablation) | multi-stream HOI baseline · V-COCO | mAP_role: 44.5 -> 46.0 | Improved |
| add contextual attention on top of context-aware appearance (ablation) | multi-stream HOI baseline · V-COCO | mAP_role: 46.0 -> 47.3 | Improved |
| add both context modules to the baseline (ablation) | multi-stream HOI baseline · V-COCO | mAP_role: 44.5 -> 47.3 | Improved |
| evaluate at a low IoU threshold (0.1) to isolate classification (ablation) | Our Approach vs baseline · V-COCO | mAP_role: 50.1 -> 53.5 | Improved |
| evaluate at a high IoU threshold (0.9) (ablation) | Our Approach vs baseline · V-COCO | mAP_role: 2.5 -> 2.8 | No clear effect |
| add context modules with VGG-16 backbone (ablation) | multi-stream HOI baseline · V-COCO | mAP_role: 42.0 -> 44.5 | Improved |
| add context modules with ResNet-101 backbone (ablation) | multi-stream HOI baseline · V-COCO | mAP_role: 45.0 -> 47.8 | Improved |
| use contextual attention approach instead of iCAN | iCAN · V-COCO | mAP_role: 45.3 -> 47.3 | Improved |
| use contextual attention approach instead of iCAN | iCAN · HICO-DET Default full | mAP: 14.84 -> 16.24 | Improved |
| use contextual attention approach instead of iCAN | iCAN · HICO-DET Known Object full | mAP: 16.43 -> 17.73 | Improved |
| use contextual attention approach instead of iCAN | iCAN · HCVRD R@50 | top-1 accuracy: 33.8 -> 37.1 | Improved |
| add context modules (inference cost) | baseline · interaction recognition time | ms per image: 111ms -> 130 milliseconds | Hurt |
Closest papers by meaning
- ContextHOI: Spatial Context Learning for Human-Object Interaction Detection (AAAI 2025 · similarity 0.92)
- Learning Human-Object Interaction Detection Using Interaction Points (CVPR 2020 · similarity 0.92)
- QPIC: Query-Based Pairwise Human-Object Interaction Detection With Image-Wide Contextual Information (CVPR 2021 · similarity 0.91)
- Learning Human-Object Interaction as Groups (NeurIPS 2025 · similarity 0.90)
- Bilateral Adaptation for Human-Object Interaction Detection with Occlusion-Robustness (CVPR 2024 · similarity 0.89)
- ContextShift: A Controlled Benchmark for Context Dependence in Object Detection (NeurIPS 2026 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
